To demonstrate that data from the Stroke and Traumatic Coma Data Banks are reliable, studies of inter-observer agreement have been implemented. These studies include a pilot study of agreement among CT Scan readers in the pilot Traumatic Coma Data Bank, and studies of variations in neurological examination, diagnosis and CT scan reading for the Stroke Data Bank. Two studies have been analyzed to date using Kappa statistics. Kappa statistics are often applied to the analysis of this type of study, yet the restrictive assumptions of this method are rarely met. An extension of the usual methods, to allow for a fixed, not random, set of raters is being developed, and guidelines for sample size (number of raters, number of subjects to be rated) are being sought for polychotomous, categorical data. Work on interobserver agreement studies is being continued under this project.